Texture-based Deep Learning for Effective Histopathological Cancer Image Classification

Tsaku, Nelson Zange; Kosaraju, Sai; Aqila, Tasmia; Masum, Mohammad; Song, Dae Hyun; Mondal, Ananda Mohan; Koh, Hyun Min; Kang, Min-Hee

doi:10.1109/bibm47256.2019.8983226

Cited by 14 publications

(9 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A patch-wise pretrained model is optimized on an objective function of a target task (e.g., classification, survival analysis) using entire patches of the training data, and each patch produces a predictive result (e.g., cancer probability) as a patch-wise analysis. In this study, we used Google-Brain (GB) 12 and CAncer-Texture Network (CAT-Net) 13 as pretrained models.…”

Section: Methodsmentioning

confidence: 99%

“…Patch-wise approaches typically consist of the three steps: (1) a WSI is divided into smaller patches; (2) each of the patches computes a probability score of a diagnosis or a clinical outcome (e.g., a probability that the patch region is cancerous); and (3) the probability scores of the patches are combined into an entire probability map of the WSI 11 . For instance, Convolutional Neural Networks (CNNs) were trained with fixed-size patch images (e.g., 299 × 299 pixels), and the patch-wise results localized cancerous regions in a WSI 12,13 . Patches in annotated Regions of Interest (ROI) were used to train a CNN-based model to predict risk scores of patches 14,15 .…”

mentioning

confidence: 99%

“…The performances of HipoMap with various top-patch size K were evaluated (Table 2). A lower value of the hyperparameter K may cause missing some important patches to conduct the slide-based analysis, whereas a higher value of K may include more false positive patches in the For the comparison, we considered two pretrained CNN models with patches of size 299×299 as a backbone model: (1) Cancer-Texture Network (CAT-Net) 13 and (2) Google Brain (GB) 12 . CAT-Net and GB were trained with the training data to classify cancer and normal slides with patches of size 299×299.…”

mentioning

confidence: 99%

See 2 more Smart Citations

Deep learning-based framework for slide-based histopathological image analysis

Kosaraju

Park

Lee

et al. 2022

Sci Rep

Self Cite

View full text Add to dashboard Cite

Digital pathology coupled with advanced machine learning (e.g., deep learning) has been changing the paradigm of whole-slide histopathological images (WSIs) analysis. Major applications in digital pathology using machine learning include automatic cancer classification, survival analysis, and subtyping from pathological images. While most pathological image analyses are based on patch-wise processing due to the extremely large size of histopathology images, there are several applications that predict a single clinical outcome or perform pathological diagnosis per slide (e.g., cancer classification, survival analysis). However, current slide-based analyses are task-dependent, and a general framework of slide-based analysis in WSI has been seldom investigated. We propose a novel slide-based histopathology analysis framework that creates a WSI representation map, called HipoMap, that can be applied to any slide-based problems, coupled with convolutional neural networks. HipoMap converts a WSI of various shapes and sizes to structured image-type representation. Our proposed HipoMap outperformed existing methods in intensive experiments with various settings and datasets. HipoMap showed the Area Under the Curve (AUC) of 0.96±0.026 (5% improved) in the experiments for lung cancer classification, and c-index of 0.787±0.013 (3.5% improved) and coefficient of determination ($$R^2$$ R 2 ) of 0.978±0.032 (24% improved) in survival analysis and survival prediction with TCGA lung cancer data respectively, as a general framework of slide-based analysis with a flexible capability. The results showed significant improvement comparing to the current state-of-the-art methods on each task. We further discussed experimental results of HipoMap as pathological viewpoints and verified the performance using publicly available TCGA datasets. A Python package is available at https://pypi.org/project/hipomap, and the package can be easily installed using Python PIP. The open-source codes in Python are available at: https://github.com/datax-lab/HipoMap.

show abstract

Section: Methodsmentioning

confidence: 99%

mentioning

confidence: 99%

mentioning

confidence: 99%

See 1 more Smart Citation

Deep learning-based framework for slide-based histopathological image analysis

Kosaraju

Park

Lee

et al. 2022

Sci Rep

Self Cite

View full text Add to dashboard Cite

show abstract

“…Multi-gigapixel whole-slide pathology images can be processed with deep learning in order to detect breast cancer [31], skin cancer [32], [33], prostate cancer [33], lung cancer [33], cervical cancer [34] and cancer in the digestive tract [35]. Some methods are even able to detect the cancer subtypes [33] or detect the spread of cancer to lymph nodes (metastasis) [36].…”

Section: A Medical and Biomedical Image Analysismentioning

confidence: 99%

Efficient High-Resolution Deep Learning: A Survey

Bakhtiarnia¹,

Zhang²,

Iosifidis³

2022

Preprint

View full text Add to dashboard Cite

Cameras in modern devices such as smartphones, satellites and medical equipment are capable of capturing very high resolution images and videos. Such high-resolution data often need to be processed by deep learning models for cancer detection, automated road navigation, weather prediction, surveillance, optimizing agricultural processes and many other applications. Using high-resolution images and videos as direct inputs for deep learning models creates many challenges due to their high number of parameters, computation cost, inference latency and GPU memory consumption. Simple approaches such as resizing the images to a lower resolution are common in the literature, however, they typically significantly decrease accuracy. Several works in the literature propose better alternatives in order to deal with the challenges of high-resolution data and improve accuracy and speed while complying with hardware limitations and time restrictions. This survey describes such efficient high-resolution deep learning methods, summarizes realworld applications of high-resolution deep learning, and provides comprehensive information about available high-resolution datasets.

show abstract

“…Therefore, it would be desirable to help pathologists more accurately determine whether a patient belongs to the EBV group only based on cost-efficient analysis of pathological images. Most recent works [6][7][8] focused on the task of gastric cancer classification into positive and negative categories, with the exception of a recent work 9 where a deep convolutional neural network (CNN) with the ResNet backbone 10 was trained to predict the molecular subtypes of gastric cancer called microsatellite instability (MSI) and microsatellite stability (MSS).…”

Section: Introductionmentioning

confidence: 99%

Self-supervised learning and multi-scale ensemble for EBV prediction based on whole slide image of gastric cancer

zhang

Wang²,

Li³

et al. 2022

Third International Conference on Computer Science and Communication Technology (ICCSCT 2022)

View full text Add to dashboard Cite

Gastric cancer is the fifth most common cancer and the fourth highest cause of cancer death in the world. One molecular subtype of gastric cancer called Epstein-Barr virus (EBV) positive tumor often responds remarkably well to immune checkpoint inhibitors and has a favorable prognosis. Considering EBV testing is often time-consuming and costly, it is of great significance to develop an automatic classification method for EBV subtype prediction based on only the cost efficient pathological images. In this study, a self-supervised learning method was proposed to train a more generalizable feature extractor (often consisting of multiple convolutional layers) using only unlabeled pathological images, based on which the classifier head (often consisting of two- or three-layer fully connected layers) can then be more efficiently trained using a large number of labelled pathological image patches. In particular, a novel formation of positive pairs for self-supervised learning was proposed by considering that neighboring patches in each pathological image often share similar visual features and therefore should have similar feature representation from output of the feature extractor. In addition, by imitating the diagnosis process of pathologists who often observe the pathological image at multiple magnifications, a multi-scale ensemble model was proposed, with each individual classifier for prediction of image patches with a unique magnification scale. Experiments on two external pathological image datasets show that the proposed self-supervised learning can help gain a more effective EBV classifier and the multi-scale ensemble model can further improve the prediction stability.

show abstract

Texture-based Deep Learning for Effective Histopathological Cancer Image Classification

Cited by 14 publications

References 10 publications

Deep learning-based framework for slide-based histopathological image analysis

Deep learning-based framework for slide-based histopathological image analysis

Efficient High-Resolution Deep Learning: A Survey

Self-supervised learning and multi-scale ensemble for EBV prediction based on whole slide image of gastric cancer

Contact Info

Product

Resources

About